WCD-02. Progress towards a global hourly-updating data assimilation system

Abstract
Currently, the US global forecast system uses a 6-hour data assimilation window, providing analyses from which to initialize forecasts every 6 hours. This is too long of a window to provide accurate information on rapidly-evolving systems such as hurricanes and continental convective storms. In addition, higher-resolution regional models need lateral boundary conditions every hour, and thus require an intermediary system to interpolate the 6-hourly global fields to hourly boundary conditions. In order to improve forecasts of these storms in both the global and regional systems, shorter assimilation windows are necessary. To this end, we are testing different approaches to updating the global forecast system every hour. A challenge for hourly updating is data latency: many observations are not available for assimilation until 1-2 hours after the valid observation time. One method to overcome this challenge is the “catch-up cycle”, in which an hourly cycling system is re-initialized 2-4 times per day from the global forecast system so that information from late-arriving observations can be injected into the hourly system via the reinitialization. Another possible solution is to implement overlapping assimilation windows, in which the system is updated every hour with observations that have arrived within the last hour, but that are valid in a longer window (3-6 hours), removing the need for a separate 6-hourly global system. The overlapping windows technique is demonstrated first in a simple setup that only assimilates conventional observations using an ensemble Kalman filter, and then in a more complicated setup that assimilates the full observing system using a hybrid gain assimilation method. Preliminary results suggest that the overlapping windows technique can improve upon short-term fit-to-observations in the GFS, relative to the current operational method.